you have to enter the transformation coordinates in the boxes under the images
Here are the steps
1) Display the image to move and click the origin bar (the small thin bar above the coordinates)
2) The mm values should be [0 0 0]
3) Click on the image until the crosshairs are where you want them
4) The values in mm are the amount you have to move the images. Let's say the values are [5 20 -10]
5) Enter the negative of the values (i.e., [0 0 0] - [5 20 -10] = [-5 -20 10]) in the right, forward and up boxes respectively.
6) Click the reorient button and select the image(s) you want to reorient. Voila they are reoriented.
7) Check the image in display.

1) Instead of clicking the image you can just enter values in the reorientation edit boxes and watch where the crosshairs will move. This is also what you will have to do to counter any rotations. (note the rotations are in radians so usually very little movement is needed).
2) You can semi-automate this by using the Util->Reorient function in batch although you will need the proper 4x4 transformation matrix.
3) You can select as many images that you want to move.

This probably occurred because the voxel size is 1/2 and so the images are shifted by that amount relative to the larger voxel sizes.


On Wed, Jul 30, 2008 at 10:50 AM, Vadim Axel <[log in to unmask]> wrote:
Hi Darren,

Thanks you lot. It looks that it might be really a cause. However, I got in trouble with this reorientation. I open one of my images and I got the picture like in stage1.jpg Then, I hit the little bar and I get the picture like on stage2.jpg (when I do the same for normalized image, the crosshairs appears somewhere at the middle). I move the crossahair back to the middle and then click on reorient  button while I select all the images I need to reorient. I see the procedure completed and the date of my hdr files is updated. However, when I click on small button again I still get crosshairs as in the stage1.jpg. What am I doing wrong?

BTW, is it possible to automate this reorientation procedure or I have to make it manually every time? Where did the problem occurred originally: in scanner or in DICOM to ANALYZE conversion?

Thank you again!

On Wed, Jul 30, 2008 at 4:36 PM, Darren Gitelman <[log in to unmask]> wrote:
My guess is that the origin of the images is not being set correctly in the small voxel images unless you normalize them. Try displaying an image and clicking the little bar above the image coordinates. This will display the image and crosshairs in relation to the actual origin. You may find that on the non-normalized images the crosshairs are not near AC-PC. If so you will need to re-orient the images so that the crosshairs are correct relative to AC-PC and then redo the stats.

On 7/30/08, Vadim Axel <[log in to unmask]> wrote:
I don't expect them to appear exactly at the right place. I am using functional localizer technique and I know in advance where more or less my region should appear. Specifically, fusiform gyrus can't appear in frontal lobe. The less preprocessing we do the better, so I prefer not to make normalization when I can localize my ROIs without it.

On Wed, Jul 30, 2008 at 3:36 PM, John Ashburner <[log in to unmask]> wrote:
If images are not spatially normalised, then there is no reason why the
activations should appear in the appropriate place on the glass brain.  What
I can't figure out is why you appeared to get the activations in the right
place when using your normal resolution data.

Best regards,

On Wednesday 30 July 2008 12:00, Vadim Axel wrote:
> Hi,
> I am scanning with the following voxel resolution: 1.56 x 1.56 x 2.4
> Unless I do normalization (voxel size: 2x2x2) I am getting the activation
> totally misplaced in glass brain (see not_normalzied screenshot; the
> activations should be in temporal  - occipital regions). Normalization
> solves the problem (see normalized screenshot). With normal resolution
> scanning (voxel 3.13 x 3.13 x 4) I get standard activations without
> normalizing data as well as with normalization.
> Any idea?
> Thanks,
> Vadim